Kdeep was trained on PDBbind database.
After docking the ligands with Vina, we inputed the resulting pose and the protein to Kdeep.
Kdeep is a convolutional neural network trained on PDBind database which predicts affinity values based
on the pose of the ligand and the protein.
Homolog based docking
Tautomers and different conformers considered with RDKit
Different conformers and tautomers were considered using RDKit.
Preparation for docking with Vina was done using MGLTools where the ligand and protein PDB files from
were converted to PDBQTs. The binding site selected was the one where the homolog had a cocrystallized ligand.
10 runs of Vina docking for each ligand.
Default Vina scoring funtion
We selected the best homolog in PDB of the provided fasta based on its e-value.
Several conformers were generated for each ligand using RDKit and tautomers were considered.
Docking runs were repeated 10 times with Vina and the best scored pose of all the runs
was selected as the final result.